Distributing data is one of the key problems in implementing efficient parallel programs. Two common distributions are BLOCK and CYCLIC, which are typically efficient for programs with nearest-neighbor and broadcast communication, respectively. Unfortunately, these distributions no longer suffice when the participating processors (nodes) are not dedicated to a parallel application. In such an environment, node activity can vary during the computation. This forces action to be taken to keep the computational load balanced, although care must be taken to ensure that the benefits of redistributing data outweigh the costs. This paper describes how to effectively handle BLOCK- and CYCLIC-based distributions when some of the nodes have at least one unrelated application process executing. For this situation, we experimentally determine redistribution cost as well as take into account variance between iterations. We also develop new variant of CYCLIC for nondedicated machines that we call GEN_CYCLIC ...

The decreasing cost of personal workstations over recent years has led to their proliferation in computing environments. Studies have shown that personal workstations are underutilized for long periods of time during regular working hours. This paper identifies this under-utilization as the idle-workstation

This article stresses the importance of capturing feedback from representative users in the early stages of product development. We present our solution to producing quality requirement specifications for radiology workstations, specifications that remain valid over time because we successfully

, Conditional Random Fields (CRFs), and Lasso amongst others. This paper describes the theory and implementation of a scalable and modular convex solver which solves all these estimation problems. It can be parallelized on a cluster of workstations, allows for data-locality, and can deal with regularizers

. The participation of volunteers will be essential to accomplish this. Don’t hesitate to contact a member of the steering committee if you would like to participate (sigapl_steering@acm.org). APL Quote Quad is again being produced on a regular basis. To recover the delay which has been accumulated in preceding years

block being assigned to one workstation of the cluster. We take advantage of the inherently regular structure of the algorithm that allows a nearly perfect static load balancing and also its locality, allowing to keep the amount of communication within a reasonable range. We obtained an acceleration

(CRFs), and Lasso amongst others. This paper describes the theory and implementation of a highly scalable and modular convex solver which solves all these estimation problems. It can be parallelized on a cluster of workstations, allows for data-locality, and can deal with regularizers such as ℓ1 and ℓ2

dimensions (e.g. the obesity epidemic), and require regular updating of anthro-pometric data collections. Aim and Objectives: This study analyzed the variation in anthropometric dimensions and body composition parameters of working women employees of different geographical zones. Material and Methods

to six over traditional PVM using regular ethernet on a cluster of workstations. 1 Introduction ...

not be as good as expected due to delays in communication. Also, the heterogeneity in communication makes it hard to take advantage, or reuse, communication strategies that were useful in regular-topology platforms, e.g., parallel machines or LAN-based clusters of workstations. For instance, broadcasting in a